.NET Authors: Lori MacVittie, Dana Gardner, Greg O'Connor, Liz McMillan, Jayaram Krishnaswamy

Related Topics: .NET, Open Source, AJAX & REA, Open Web, Cloud Expo

.NET: Blog Feed Post

For Thirty Pieces of Silver My Product Can Beat Your Product

Computing hardware suffered the steepest spending decline of the four major IT spending category segments in 2009

One of the side-effects of the rapid increases in compute power combined with an explosion of Internet users has been the need for organizations to grow their application  infrastructures to support more and more load. That means higher capacity everything – from switches to routers to application delivery infrastructure to the applications themselves.  Cloud computing has certainly stepped up to address this, providing the means by which organizations can efficiently and more cost-effectively increase capacity. Between cloud computing and increasing demands on applications there is a need for organizations to invest in the infrastructure necessary to build out a new network, one that can handle the load and integrate into the broader ecosystem to enable automation and ultimately orchestration.

Indeed, Denise Dubie of Network World pulled together data from analyst firms Gartner and Forrester and the trend in IT spending shows that hardware is king this year.

blockquote"Computing hardware suffered the steepest spending decline of the four major IT spending category segments in 2009. However, it is now forecast to enjoy the joint strongest rebound in 2010," said George Shiffler, research director at Gartner, in a statement.

That is, of course, good news for hardware vendors. The bad news is that the perfect storm of increasing capacity needs, massively more powerful compute resources, and the death of objective third party performance reviews result in a situation that forces would-be IT buyers to rely upon third-parties to provide “real-world” performance data to assist in the evaluation of solutions. The ability – or willingness - of an organization to invest in the hardware or software solutions to generate the load necessary to simulate “real-world” traffic on any device is minimal and unsurprising. Performance testing products like those from Spirent and Ixia are not inexpensive, and the investment is hard to justify because it isn’t used very often.  But without such solutions it is nearly impossible for an organization to generate the kind of load necessary to really test out potential solutions. And organizations need to test them out because they, themselves, are not inexpensive and it’s perfectly understandable that an organization wants to make sure their investment is in a solution that performs as advertised. That means relying on third-parties who have made the investment in performance testing solutions and can generate the kind of load necessary to test vendor claims.



That’s bad news because many third-parties aren’t necessarily interested in getting at the truth, they’re interested in getting at the check that’s cut at the end of the test. Because it’s the vendor cutting that check and not the customer, you can guess who’s interests are best served by such testing.


Now generally speaking such third-party “reviews” (and I use that term loosely) are purported to be “comparative”. But what they really are is competitive; they are specifically fingerscrossed2designed (paid for) to be used by sales folks in competitive situations. You can tell these tests aren’t really comparative because they purport to be testing the same product (in the sense that they’re designed to do the same thing) but one of the two products being “compared” turns out to be less capable in the speeds and feeds department. The product which the test is designed to highlight will be rated for 2GBps of throughput while the competitive platform will be rated for only 1GBps. It’s like setting up a race between a 4 year old and a 2 year old. Guess which one is going to win? (I’d put my money on the 4 year old, by the way, and on the higher-throughput rated device. Longer legs, you see.)

It’s not that these tests aren’t valuable. They are – for the product of the vendor who cut the check. The performance results for the product being highlighted are almost certainly valid, having been tested under the watchful eye of the commissioner of the test and tweaked by experts who know how to squeeze out every last gigabit of performance (don’t forget to ask if the services of experts is included in the purchase price of $product). But for the “compared” products? Not so useful. Such tests often claim to have had the competitive products’ configuration verified by a certified $vendor expert. That sounds impressive, until you discover that $vendor doesn’t have any “certified” experts. Worse, the “certified” experts aren’t affiliated with $vendor because the invitation to participate came a week before the test was being concluded, giving $vendor no time in which to find the appropriate expert and allow them to travel to assist. It’s a purposefully deceitful practice that, in my days of comparative testing, would have been loudly protested. Two months before testing begins, not two weeks before testing ends, was the rule for sending invitations then, back when comparative meant comparative and not merely competitive.

Of course we weren’t paid for our services, either. We were paid to evaluate products and provide all the details for IT purchasers so they had the data necessary to help them compile a short list of $vendor products to test themselves. We were not in the business of producing a competitive document that could be used to convince customers of the superiority of $product at any cost.


An interesting side-effect of paying for a competitive test is that if you’re the one paying for it, you get to decide what data goes into the report. Don’t like how one of the performance tests came out? Don’t include it. Handily beat the competitor in CPU utilization even though without detailing the test scenario such a comparison is irrelevant? Don’t just include it, make sure it’s a major point of “comparison”. Using a different definition for a performance metric than the rest of the industry because it makes $product appear to perform better? Highlight the result, but do not under any circumstances include your definition of that performance metric. Even better is to make up a performance metric; I’m partial to gigafluxes per second, which is a completely made up metric but sounds like it might be valid and thus might fool enough potential customers into purchases to pay for the cost of the test. If you see it in the wild, drop me a line; you know I always enjoy a good laugh.

In one recent paid test, a $20K 1Gbps application delivery controller was competitively tested against a 4Gbps Load-Balancer selling for over $50K.  Guess what, the $50K Load-Balancer was faster.  But you would never know that from the report because nothing in the report indicates that the two devices were of two completely different classes in terms of performance and functionality. That’s why this post includes ten objective questions you should ask any vendor when presented with “independent” competitive test reports. Integrity is not something you buy, it’s something you earn in the process of helping customers make informed decisions.

If the vendor who paid for the test cannot answer any one of these questions adequately, you should probably read the report very skeptically and treat it simply as another sales tool for the vendor. You shouldn’t read it as a tool for making a good product choice.

I could go on and on about all the ways in which these competitive tests are intentionally obfuscating the truth but I’ve already done that years ago in “Vendorspeak Exposed!” and the marketing tactics upon which that lengthy diatribe was based haven’t changed one whit, except that perhaps because of circumstances vendors are now forced to rely more than ever on third-party pay-for-play competitive testing. But the same tactics – false precision, deceptive eye-candy, misleading metric definitions, incomplete configuration and test-bed documentation – are still used by those third-party firms. And I can’t even blame them for doing so because they have to pay bills, too. But at least they could be honest about it and not pretend that they conduct comparative testing when they’re really providing competitive reports.

What it comes down to is that fairly executed, comparative product reviews are dead. But as Mike Rothman commented in his lament, “The Death of Product Reviews”, folks in the trenches still need information and guidance in making product decisions.

blockquote What now for the folks in the trenches? Once the hangover from the wake subsides, we still need information and guidance in making product decisions. So what to do? That's a topic for another post, but it has to do with structuring the internal proof of concept tests to reflect the way the product will be used -- rather than how the vendor wants to run the test.

-- Mike Rothman, “The Death of Product Reviews

So what’s a poor IT trencher to do? Data sheets from $vendor provide performance data, yes. But that data is “best case scenario”. It should be read as “this is the best performance you will see out this product, YMMV.” Mike’s suggestion to structure internal proof of concept tests to reflect real-world usage is a good one, though it may be difficult to implement in practice (we talked about the cost factor earlier, which makes internal POC testing a tad cost-prohibitive). The rise of cloud computing may actually be a boon for organizations in this regard. The availability of on-demand compute resources that can be leveraged to generate load for testing real-world capacity and performance limitations makes it more possible today than it ever was for organizations to generate the kind of load necessary to determine capacity based on their real-world usage. Even better are cloud-based load testing services like SOASTA and LoadImpact that have the foundation for test harnesses built into their platforms and can generate load against devices and sites remotely in what is certainly more real-world than a closed, sterile lab environment. The latter is better, in my opinion, because it stresses the entire infrastructure just as it would be stressed under real load. Just don’t kick off a load test during business hours – you don’t want to find a bottleneck in your network by DoSing it when folks are trying to work.

I’m not saying ignore competitive reports when they’re offered. In fact, they can offer you some insight into what questions you should be asking all the vendors involved in your acquisition process. I am saying that while the veracity of the report with regards to the $vendor product being referenced is likely trustworthy, the rest of the report probably needs to be viewed through a lens of skepticism and with the understanding that it is a competitive report, not a comparative one.

After all, would you pay upwards of $40,000 for a report in which your product didn’t come out on top?

Related blogs & articles:

Follow me on Twitter View Lori's profile on SlideShare friendfeed icon_facebook

AddThis Feed Button Bookmark and Share

Technorati Tags: ,,,,,,,,,,

Read the original blog entry...

More Stories By Lori MacVittie

Lori MacVittie is responsible for education and evangelism of application services available across F5’s entire product suite. Her role includes authorship of technical materials and participation in a number of community-based forums and industry standards organizations, among other efforts. MacVittie has extensive programming experience as an application architect, as well as network and systems development and administration expertise. Prior to joining F5, MacVittie was an award-winning Senior Technology Editor at Network Computing Magazine, where she conducted product research and evaluation focused on integration with application and network architectures, and authored articles on a variety of topics aimed at IT professionals. Her most recent area of focus included SOA-related products and architectures. She holds a B.S. in Information and Computing Science from the University of Wisconsin at Green Bay, and an M.S. in Computer Science from Nova Southeastern University.